From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <glibc-bugs-return-18253-listarch-glibc-bugs=sources.redhat.com@sourceware.org>
Received: (qmail 26167 invoked by alias); 27 Mar 2013 21:13:47 -0000
Mailing-List: contact glibc-bugs-help@sourceware.org; run by ezmlm
Precedence: bulk
List-Id: <glibc-bugs.sourceware.org>
List-Subscribe: <mailto:glibc-bugs-subscribe@sourceware.org>
List-Post: <mailto:glibc-bugs@sourceware.org>
List-Help: <mailto:glibc-bugs-help@sourceware.org>, <http://sourceware.org/lists.html#faqs>
Sender: glibc-bugs-owner@sourceware.org
Received: (qmail 26118 invoked by uid 48); 27 Mar 2013 21:13:40 -0000
From: "law at redhat dot com" <sourceware-bugzilla@sourceware.org>
To: glibc-bugs@sourceware.org
Subject: [Bug dynamic-link/15310] _dl_sort_fini is O(n^3) causing slow exit when many dsos
Date: Wed, 27 Mar 2013 21:13:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: glibc
X-Bugzilla-Component: dynamic-link
X-Bugzilla-Keywords:
X-Bugzilla-Severity: critical
X-Bugzilla-Who: law at redhat dot com
X-Bugzilla-Status: WAITING
X-Bugzilla-Priority: P2
X-Bugzilla-Assigned-To: unassigned at sourceware dot org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Changed-Fields:
Message-ID: <bug-15310-131-z4YfQDgFta@http.sourceware.org/bugzilla/>
In-Reply-To: <bug-15310-131@http.sourceware.org/bugzilla/>
References: <bug-15310-131@http.sourceware.org/bugzilla/>
X-Bugzilla-URL: http://sourceware.org/bugzilla/
Auto-Submitted: auto-generated
Content-Type: text/plain; charset="UTF-8"
MIME-Version: 1.0
X-SW-Source: 2013-03/txt/msg00152.txt.bz2

http://sourceware.org/bugzilla/show_bug.cgi?id=15310

--- Comment #7 from law at redhat dot com 2013-03-27 21:13:39 UTC ---
Carlos,

This is the tester QE uses for the correctness of the sort:  

Basically it takes as an input the number of DSOs you want to test.  Then
builds a random orderings and a pathological ordering of those DSOs on the link
line (the script knows the dependencies between DSOs).
deps.cpp:

#include <stdio.h>

#ifdef MAIN
int
main( int argc, char *argv[] )
{
  printf("main\n");
}
#else
static bool deps_init( void ) { puts( NAME ); }
bool x = deps_init();
#endif


driver:
#!/bin/bash
# based on reproducer from bug 11724

N=${1:-2}
TmpDir=$(mktemp -d)
cp deps.cpp $TmpDir
pushd $TmpDir > /dev/null 2>&1

for((all=0,k=2; k<=$N; ++all,++k)); do
    cmd="g++ -fPIC -shared -o libdeps${k}.so -DNAME=\\\"${k}\\\" deps.cpp"
    printf "# %s\n" "$cmd" > ${all}r.log
    eval "$cmd"
    for ((i=k-1; i>=1; --i)); do
        cmd="g++ -L. -fPIC -shared -o libdeps${i}.so -ldeps$((i+1))
-DNAME=\\\"${i}\\\" deps.cpp"
        printf "# %s\n" "$cmd" >> ${all}r.log
        eval "$cmd"
    done; unset i
    # random link order
    # it sucks that rhel5 coreutils don't have shuf or sort -R yet :)
    roptions=($(seq 1 $k | sed 's/^/ -ldeps/'))
    for ((i=(${#roptions[@]}-1); i >= 1; --i)); do
        random=$((RANDOM % i))
        m=${roptions[$random]}
        roptions[$random]=${roptions[$i]}
        roptions[$i]=$m
    done; unset i
    rcmd="g++ -Wl,-rpath -Wl,$(pwd) -L. -o rmain ${roptions[@]} -DMAIN
deps.cpp"
    # pathological link order
    poptions=($(seq $k -1 1 | sed 's/^/ -ldeps/'))
    pcmd="g++ -Wl,-rpath -Wl,$(pwd) -L. -o pmain ${poptions[@]} -DMAIN
deps.cpp"

    cp ${all}r.log ${all}p.log

    for i in r p; do
        cmd="${i}cmd"
        printf "# %s\n\n" "${!cmd}" >> ${all}${i}.log
        eval "${!cmd}"

        fail=0
        cp ${all}${i}.log ${i}out.golden
        ./${i}main >> ${all}${i}.log
        seq $k -1 1 >> ${i}out.golden
        echo main >> ${i}out.golden
        diff ${all}${i}.log ${i}out.golden > /dev/null 2>&1 || fail=1

        # it's ok, we don't care
        if [ 0 -eq $fail ]; then
            rm ${all}${i}.log
        fi 
    done; unset i

    rm -rf *.so *main *out.golden *out.real
done; unset k

rm deps.cpp
test 0 -eq $(ls | wc -l) && { rm -rf $(pwd); echo PASS; } || { echo FAIL;
printf "results are in %s\n" $(pwd); }
popd > /dev/null 2>&1

-- 
Configure bugmail: http://sourceware.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.